Re: Use of rtems_fdt_* and sp01

[Joel Sherrill]

On Sat, Nov 19, 2022, 4:19 PM Chris Johns  wrote:

> On 19/11/22 2:13 am, Joel Sherrill wrote:
> > On Fri, Nov 18, 2022, 8:49 AM Kinsey Moore  > > wrote:
> > On Fri, Nov 18, 2022 at 12:44 AM Sebastian Huber
> >  > > wrote:
> > On 18/11/2022 06:32, Kinsey Moore wrote:
> > > On Thu, Nov 17, 2022 at 4:49 PM Chris Johns  > 
> > > >> wrote:
> > >
> > > On 18/11/2022 2:39 am, Kinsey Moore wrote:
> > >  > I recently added FDT support to the AArch64 ZynqMP BSPs
> to
> > > support an optional
> > >  > management console and managing ethernet parameters for
> LibBSD.
> > > Use of the
> > >  > rtems_fdt_* functions implies use of malloc which adds
> 4 bytes in
> > > the TLS space.
> > >  > The sp01 test uses rtems_task_construct and specifies a
> maximum
> > > TLS size of 0
> > >  > bytes which causes a failure which the non-zero TLS
> size is
> > > checked against the
> > >  > maximum TLS size of 0.
> > >
> > > Does this mean there exists a requirement a BSPs cannot
> internally
> > > use the heap?
> > >
> > >
> > > That appears to be the case, at least practically. It works
> fine, but it
> > > causes a test failure...
> >
> > You can use the heap during system initialization. However, you
> should
> > avoid dependencies on errno, so instead of using
> malloc()/calloc() use
> > rtems_malloc()/rtems_calloc(). Independent of this, if the BSP
> > initialization can easily avoid the heap, then it should not use
> the heap.
> >
> > >
> > >
> > > Maybe the test needs to be adjusted?
> > >
> > >
> > > That's part of why I sent this to the list. I was unsure
> whether it was
> > > allowed or whether the test had bad assumptions baked into it.
> >
> > The tests check specific aspects of the thread-local storage
> support and
> > this works only if the BSP does not depend on TLS objects such
> as errno.
> >
> > This affects several other tests as well, so I guess my solution
> here is
> > either to alter the rtems_fdt_ wrappers to avoid dependencies on TLS
> or
> > proceed with the existing patch that uses the non-wrapped FDT
> functions.
> >
> > I agree with Sebastian that fixing the bsp is right.
> >
> > But also rtems_ fdt methods should be changed to use rtems_malloc. That
> would
> > leave those methods useful. They are broken now for many intended use
> cases.
>
> We are currently reviewing the support for ftbo files in Linux (petalinux)
> with
> a view to seeing how we can support them on RTEMS. I see FDT being use
> more and
> more in RTEMS and with a number of different use cases and flows there is
> a need
> to try and unify how we handle them. I see benefits if we can support a
> similar
> dtbo model as Linux and if that happens the specific area in question in
> rtems-fst can be move over to using it.
>

FWIW Kinsey's patch to the rtems_fdt methods eliminated two coverity issues.

>
> Chris
>
___
devel mailing list
devel@rtems.org
http://lists.rtems.org/mailman/listinfo/devel

Re: Use of rtems_fdt_* and sp01

[Chris Johns]

On 19/11/22 2:13 am, Joel Sherrill wrote:
> On Fri, Nov 18, 2022, 8:49 AM Kinsey Moore  > wrote:
> On Fri, Nov 18, 2022 at 12:44 AM Sebastian Huber
>  > wrote:
> On 18/11/2022 06:32, Kinsey Moore wrote:
> > On Thu, Nov 17, 2022 at 4:49 PM Chris Johns  
> > >> wrote:
> >
> >     On 18/11/2022 2:39 am, Kinsey Moore wrote:
> >      > I recently added FDT support to the AArch64 ZynqMP BSPs to
> >     support an optional
> >      > management console and managing ethernet parameters for 
> LibBSD.
> >     Use of the
> >      > rtems_fdt_* functions implies use of malloc which adds 4 
> bytes in
> >     the TLS space.
> >      > The sp01 test uses rtems_task_construct and specifies a 
> maximum
> >     TLS size of 0
> >      > bytes which causes a failure which the non-zero TLS size is
> >     checked against the
> >      > maximum TLS size of 0.
> >
> >     Does this mean there exists a requirement a BSPs cannot 
> internally
> >     use the heap?
> >
> >
> > That appears to be the case, at least practically. It works fine, 
> but it
> > causes a test failure...
> 
> You can use the heap during system initialization. However, you should
> avoid dependencies on errno, so instead of using malloc()/calloc() use
> rtems_malloc()/rtems_calloc(). Independent of this, if the BSP
> initialization can easily avoid the heap, then it should not use the 
> heap.
> 
> >
> >
> >     Maybe the test needs to be adjusted?
> >
> >
> > That's part of why I sent this to the list. I was unsure whether it 
> was
> > allowed or whether the test had bad assumptions baked into it.
> 
> The tests check specific aspects of the thread-local storage support 
> and
> this works only if the BSP does not depend on TLS objects such as 
> errno.
> 
> This affects several other tests as well, so I guess my solution here is
> either to alter the rtems_fdt_ wrappers to avoid dependencies on TLS or
> proceed with the existing patch that uses the non-wrapped FDT functions.
> 
> I agree with Sebastian that fixing the bsp is right.
> 
> But also rtems_ fdt methods should be changed to use rtems_malloc. That would
> leave those methods useful. They are broken now for many intended use cases.

We are currently reviewing the support for ftbo files in Linux (petalinux) with
a view to seeing how we can support them on RTEMS. I see FDT being use more and
more in RTEMS and with a number of different use cases and flows there is a need
to try and unify how we handle them. I see benefits if we can support a similar
dtbo model as Linux and if that happens the specific area in question in
rtems-fst can be move over to using it.

Chris
___
devel mailing list
devel@rtems.org
http://lists.rtems.org/mailman/listinfo/devel